Skip to main content

Full text of "The interrelation of yield and protein content of random selections from single crosses in wheat and barley"

See other formats


■HHI PM 




For Reference 


NOT TO BE TAKEN FROM THIS ROOM 


THE INTERRELATION OP YIELD AND PROTEIN CONTENT 
OF RANDOM SELECTIONS FROM SINGLE CROSSES 
IN WHEAT AND BARLEY 


M. N. Grant 

Department of Plant Science 


University of Alberta 




w wmm / ., 

jkmm V im- Vv • ‘v«> - k' 




v>. j- ; V V V J ' \ t ' )' i $ v V . 


y.vu* m 




<-•- c ■ a. 


(11*1 





















































Ox UB8M 

mwibswmis 

attmiiwis 







Digitized by the Internet Archive 
in 2018 with funding from 
University of Alberta Libraries 


https://archive.org/details/interrelationofyOOgran 





THE INTERRELATION OF YIELD AND PROTEIN CONTENT 
OF RANDOM SELECTIONS FROM SINGLE CROSSES 
IN WHEAT AND BARLEY 


M. N. Grant 

Department of Plant Science 


A THESIS 

submitted to the University of Alberta 
in partial fulfilment of the 
requirements for the degree of 
MASTER OF SCIENCE 


Edmonton, Alberta 
April, 1946 






















I 








TABLE OF CONTENTS 

Page 

Introduction .. 1 

Literature Review .. 3 

Material and Methods ... 8 

Experimental Results . 16 

Barley . 16 

1944 . 16 

1945 . 20 

Means of two tests . 27 

Xnteryear .. 27 

Wheat .. 32 

1944 . 32 

Edmonton • ..... • 32 

Fallis . 36 

1945 . 39 

Edmonton . 39 

Fallis . 43 

Means of four tests . 45 

Interstation .. 50 

Interyear . 55 

Discussion . 62 

Summary.*.... 66 

Acknowledgements . 68 

References . 68 




















































THE INTERRELATION OF YIELD AND PROTEIN CONTENT 
OF RANDOM SELECTIONS FROM SINGLE CROSSES 
IN WHEAT AND BARLEY 


M. N. Grant 


INTRODUCTION 


Yield and protein content are the most important 
characters in Canadian wheat and barley, the former because of 
its direct effect on the farmer’s revenue, and the latter 
because of its relation to baking quality in wheat and malting 
and feed value in barley. The inheritance of both these 
characters is undoubtedly controlled by a number of genetical 
and environmental factors. Of the two, yield is the more dif¬ 
ficult to analyze since, as stated by Worzella (20), it may be 
regarded as the ultimate expression of all the inherent factors 
and environmental conditions that have been associated through¬ 
out the life of the plant. The analysis of yield can best be 
carried out by the identification of its component parts, and 
by a study of their nature, interrelationships, and behavior 
under varying environmental conditions. The plant breeder is 
interested primarily in the inherent factors affecting yield, 
since they represent the components that can be permanently 
modified by breeding methods. One of these components is pro¬ 


tein content 





\ 

■ 1 ■ 

.■' 











- 2 - 


Examination of data from many sources suggests a 
definite negative association between yield and protein content 
of wheat and of barley, part of which association may be attri- 
buted to a linkage between the genes for the two characters® 

It is extremely important that a measure of this association 
be obtained, since, if it can be proved generally true that 
varieties which are characterized by high yield tend to be low 
In protein, then the problem of Improvement of malting barley 
and soft wheats by breeding methods would be greatly simpli¬ 
fied® In the case of hard red spring wheats or feed barley, 
however, where both high yield and high protein content are 
desired, the attainment of this end is made much more difficult 
by a genetical relationship such as has been suggested* 

This relationship also suggests a simple specific 
test for the elimination or selection of hybrid strains in 
early generations of wheat and barley* The problem has been 
approached by many workers (7, 8, 10, 11, 17) who have 
attempted to find a means of saving time ai d work in the selec¬ 
tion of Improved strains from early-generation hybrid material* 
The importance of determining a simple means for such a test 
cannot be over-emphasized, since it is relatively easy to make 
many crosses which will result in thousands of new genotypes; 
but It is much more difficult, in the large mass of material, 
to recognize which strains are valuable and which, are worthless. 
If the negative association between yield and protein content 


: " ' ■ 

■■■; ' 

• ■ ■ •• . ' -v-r 

■ r," ! . 

-: ! I " 

* 

, ' 

‘ • : 1 " V 1 

1 ' ■ ' ' & 

■ 






3 - 


could be proved true in a sufficiently high proportion of the 
cases studied, many strains could be discarded in early genera¬ 
tions on the basis of a protein determination, thus saving 
years of tedious work while these strains were increased to a 
point where enough seed had become available for yield trials. 

The purpose for which the following investigation 
was initiated was to endeavor to establish definitely the 
relationship which was believed to exist between yielding 
ability and protein content in wheat and barley. 


LITERATURE REVIEW 


The negative relationship between yield and protein 
content has attracted the attention of many workers. Reference 
to this phenomenon is found in many publications, but unfortun¬ 
ately the problem has, in almost every case, been regarded as 
an interesting phase of some other project, and has not been 
investigated thoroughly itself. It is surprising to note the 
small amount of experimental work which has been carried out 
on this useful and generally accepted correlation between yield 
and protein content. 

In 1926, Goulden and Elders (6) pointed o\it that, in 
wheat improvement, the primary objectives are yield and quality. 
Yield, in their estimation end in that of Clark and Quisenberry 
(3), Is a complicated character, determined by a series of 



- ■ 
















■ 








t 











4 


lesser physiological and morphological characters. Clark ( 2 ) 
and McCalia and Rose (13) state that the most reliable indi¬ 
vidual measure of quality in wheat is the crude protein con¬ 
tent, which can be determined with a high degree of accuracy 
on a small quantity of grain. In an F g population of Marquis x 
Hard Federation wheats, Clark (2) found that the crude protein 
content of the hybrid plants was negatively correlated with 
yield, the coefficient of correlation being -.231 1 .027. 

Yield affected the protein content more than any other charac¬ 
ter studied. These characters included dates of heading and 
ripening, fruiting period, and height. In two tests conducted 
on the plants the correlation coefficients were, respectively, 
+.256 * .042 and -.732 t .018. Clark concludes that there is 
segregation for crude protein content in wheat hybrids similar 
to that for other quantitative characters, including yield. 
Inheritance of crude protein content is as complex as that of 
yield, and environment has as much effect on the one as on the 
other. The two characters are frequently, but not always, 
negatively associated. 

Clark and Quisenberry (3), in a study of the yield 
and protein content of 181 Fg plants from a Kota x Marquis 
cross, concluded that there is a slight but not significant 
tendency for the two characters to be negatively correlated. 

They did find, however, a significant positive correlation 
between the crude protein content of Fg plants and F 5 strains, 
indicating that, in breeding for high protein content, the 







>s 






selection of high-protein plants in the segregating Fg genera¬ 
tion offers a promising method of attack, 

A different approach to the problem was made by 
Malloeh and Newton (12) who tested the relationship between 
yield and protein content of single varieties as affected by 
variations in the soil and in pruning the plants. Fifty rod- 
row samples were selected at random and analyzed for yield and 
protein content. The correlation coefficient was -«68 for Red 
Bobs in 1930, and -.42 for Marquis in 1931, It was also shown 
that a reduction in yield by pruning resulted in an increase 
in protein content of the grain. Their conclusion was that 
there is usually an inverse relation between yield and protein 
content, though it is not so definite as to mean that yield in 
all cases can be increased only at the expense of a decrease 
in the protein content. 

In a study of yield and other characters of 25 varie¬ 
ties of hard red spring ivheat, Waldron (18) obtained a correla¬ 
tion coefficient of -,556 for yield and protein content, with 
a regression of yield on protein content of -3,4 bushels* 
Waldron stresses the economic importance of this relationship, 
indicating that with a. regression as great as this a distinct 
protein premium on the market would be necessary to recompense 
tlie farmer for his high-protein wheat, since the yield of his 
crop would be decreased* 

Whiteside (19) states that in a study of 28 varieties 




5 ! 







- 6 - 


of spring wheat grown at each of three stations, yield and 
protein content were not correlated when the effects of sta¬ 
tion, replication, and variety were removed. No real differ¬ 
ences were found between the calculated protein percentages 
for composite samples made up from the four plots from each 
station and the percentages obtained by averaging the results 
from the individual plots. The two sets of protein percentages 
gave a correlation coefficient of +.9998. 

Neatby and McCalla (15) examined yield and protein 
data secured from hard wheat, soft wheat, and barley variety 
trials conducted over a period of years at the University of 
Alberta, together with data obtained on numerous tests in 
Saskatchewan, Manitoba, Washington, Oregon, and Utah. For 1933 
the correlation between the general means of yield and protein 
for 36 varieties of hard wheat grown at 11 stations was deter¬ 
mined. This coefficient, -.55, is almost certainly due largely 
to genetical causes. When samples of winter wheat from five 
stations were composited for protein determinations, a yield- 
protein correlation coefficient of -.57 was obtained. Working 
with soft wheat data, the general means of yield and protein 
for 11 varieties at 7 stations were correlated and gave an 
extremely high correlation coefficient of -.92. A similar 
procedure applied to barley data indicated that the relation 
holds also for this crop. The correlation coefficient for the 
means of yield and protein in this case was -.72. It was con- 











7 


eluded that yield and protein content in these cereals have 
a genetically controlled negative association with each other* 
Yielding ability and protein content are assumed to be con¬ 
trolled by the same laws of inheritance as are other more 
obvious characters, subject of course to environmental influ¬ 
ences such as moisture, soil nitrogen, and general nutritional 
conditions* This supports the conclusions of Clark (2)* The 
constancy of these genetical differences for both yield and 
protein content when the effect of environment is controlled 
was demonstrated* 

McCalla and Rose (13) state that high-protein 
varieties of wheat are almost invariably low in yield, and 
suggest that great difficulty would be experienced in an attempt 
to obtain a high-yielding, high-protein, hard spring wheat* 

In their cultural studies with barley, Olson, Meredith, 
Laidlaw, and Lejeune (16) state that the important objectives 
from the point of view of both producer and consumer are yield 
and malting quality* Meredith and Olson (14) point out the 
importance of protein content in the evaluation of malting 
quality* The desirability of a low-protein malting barley is 
stressed again by Anderson, Meredith, and Sallans (1). 
















* 










* 


- 8 - 


MATERIAL ALT) METHODS 


1944 


Barley 

The cross, Trebi x Featland, was made in 1936 to pro- 
yide material for a genetical study of the relationship between 
yield and protein content. Peatland is characterized by very 
high protein content and low yield; while Trebi is consistently 
a high-yielding, low-protein variety. The seed of the cross 
was grown in bulk plots where it was allowed to increase and 
segregate till the fall of 1939 when 100 random selections 
were made of plants carrying seed. 

It was thought that, if the selection of heads was 
made at random, and if yielding ability and protein content are 
Inherited independently, the samples would show equal numbers 
of high-yield-high-protein, low-yield-low-protein, high-yield- 
low-protein, and low-yield-high-protein selections, plus 
intermediate forms. If, however, the scatter of points obtained 
by plotting yield in bushels per acre against protein content 
showed any definite trend, this information might be of value 
In carrying out future selection work with hybrid populations• 

An explanation of how this information could be used is illus¬ 
trated by the correlation surface in Figure 1. 



















PROTEIN % 


- 9 - 



YIELD , £ 4 /. 


Figure 1 

Correlation surface illustrating general 
association between yield and 
protein content 



























I 


- 10 


If the association shown in Figure 1 could be proved 
to be generally true, it would then be possible to use a pro¬ 
tein determination as a means of selection with small quanti¬ 
ties of hybrid seed. The extremely high protein selections 
could be discarded since they would, in all probability, 
later prove to be low in yield. If the plants under test were 
destined to be feed barleys, then those with very low protein 
content could be discarded for that characteristic alone. 

In short, for feed barleys, the most desirable segregates are 
found in quadrant 1 , while for malting barleys the best 

4 

selections would be plotted in quadrant 2 . If the association 
between yield and protein content were known, protein analysis 
alone could then afford a sufficiently accurate method of 

selection. 

The seed from the 100 plants selected in 1939 was 
planted in separate rows and Increased during 1940. The fol¬ 
lowing spring the 100 F 5 selections were placed in a quadrupli¬ 
cated yield test. Lack of sufficient moisture during the grow¬ 
ing season resulted in a poor stand for all selections. The 
yield results in particular were not thought to be as accurate 
as would be desired for correlation studies with protein con¬ 
tent. The analysis of the protein content of each selection 
was carried out, however, and the results recorded. Calcula¬ 
tions based on the 1941 data give a regression coefficient for 
yield and protein content that is negative and highly signifi¬ 
cant (bpy «* -.037). The regression equation Is P = 17.27 - 





• > J ; 








- 11 


*037 y* In Figure 2 is presented the scatter of points and 
the regression line obtained for this test. The simple cor¬ 
relation coefficient Is not large but is highly significant 

(r?y * -.300)• 

In 1944 a yield trial was again grown at Edmonton 
using seed from the 1940 increase plots. Only 82 selections 
were included, using a simple randomized block design with 
four replicates* Replicates were splic to form more compact 
and uniform units. Each plot consisted of three rows, 18.5 
feet long, and 9 inches apart, sown at the rate of 2 bushels 
per acre. Seeding was done with a V-belt seeder. No ferti¬ 
lizer was added. 

Field notes were taken on heading date and growth 
period (days from seeding to maturity). At maturity the 
centre row of each plot was trimmed to 16.5 feet to minimize 
any border effect from the open pathways. Each centre row 
was cut by hand and wrapped in a cotton wrapper to prevent 
mixing and loss of seed. Each sheaf was threshed in a rod- 
row thresher, and the yield recorded. Protein analysis was 
carried out on composite samples of seed from the four repli¬ 
cates, a method considered satisfactory by Whiteside (19). 

Analysis of variance was applied to the yield but 


not to the protein data 


, z 


















< 












- 12 



YIELD , BU. 


Figure 2 

Relation between yield and protein content 
in the 1941 barley test 












I 






I 










• I 











Wheat 


The cross selected for study of the yield-protein 
relationship was Bunyip x Dicklow. Both parents are soft 
wheats. Bunyip is characterized by high protein and low yield, 
while Dicklow is noted for low protein and high yield. The 
hybrid seed was grown in bulk plots till the P 4 generation in 
1940 when random selections of plants were made. These selec¬ 
tions were increased during 1941. Yield trials were carried 
out in 1944 at both Edmonton (black soil) and Fallis (gray 
soil) using seed of 100 selections from the increase plots of 
1941. 

The field plot design was of the incomplete block 
type, known as the simple lattice, and illustrated by Cox and 
Eckhardt (4) and Hayes and Immer (9). Since the number of 
selections forms a perfect square and v = k 2 , v (varieties) 
equals 100, and k (blocks) equals 10 . The 100 selections were 
arranged in incomplete blocks, each block containing 10 selec¬ 
tions. There were four replicates, each made up of 10 blocks. 
The improved efficiency of the lattice design over simple ran¬ 
domized blocks has been established where large numbers of 
selections are to be used, as shown by Cox and Eckhardt (4). 

These plots, like those for the barley, consisted of 
3 rows, 18.5 feet long and 9 inches apart, but were sown at the 
rate of 1.25 bushels per acre. No fertilizer was added at 


either station 




. 




. 

. 

• f . 

* 















— 14 «■* 


Field notes were taken at Edmonton, but not at Fallis, 
on heeding date and growth period. The centre row of each plot 
was trimmed to 16.5 feet at maturity, cut by hand, wrapped, and 
threshed separately in a rod-row thresher. The yield was 
recorded. 

The analysis of variance for yield data was applied 
in the manner outlined by Hayes and Immer (9). 

Protein content was determined on composite samples 
of seed from the four replicates. No analysis of variance, 
therefore, could be applied. 

1945 


Barley 

The yield trial, with certain changes, was repeated 
in 1945. Seed was obtained from the 82 Fg selections in the 
1944 test, and to these were added 18 Fg selections from the 
1941 test. This increased the number of selections to 100, thus 
making possible the use of the simple lattice design of Cox 
and Eckhardt (4), which had proved satisfactory for the vrbeat 
tests in the previous year. The seeding methods were the same 
as those outlined for 1944. 

At harvest time it ras decided that the seed from 
each plot nvould be bagged separately, and that protein content 






' 








' 





15 


would be determined for each sample* This made possible the 
use of analysis of variance methods for the protein as well 
as the yield and growth, period data, and also gave the 
opportunity to apply a correction factor to the mean values 
for each selection before correlation studies were attempted. 

Wheat 


The two wheat tests were repeated in 1945, one on 
the black soil at Edmonton, and the other on the gray soil at 
FalliSo The seed was taken from the selections which had 
been grown during 1944. 

At harvest the seed from each plot was placed in 
individual bags, and protein content determined for each sample. 
Analysis of variance methods were applied to the yield, pro¬ 
tein, and growth period data, and correction factors were used 
in every case before correlation studies were begun. 



















16 - 


EXPERIMENTAL RESULTS 


Barley 


1944 


For the months of May, June, and July, the total 
precipitation exceeded 11 Inches* With this ample supply of 
moisture all the barley selections exhibited a heavy growth* 

At heading time the selections showed notable differences in 
height, straw strength, and earliness, indicating the segrega¬ 
tion which these characters and presumably others, including 
yielding ability and protein content, had undergone during the 
preceding generations* The agronomic differences are apparent 
in Figure 3* 

The results of the analysis of variance for yield and 
for growth period are presented in Table I* In the case of 
yield highly significant differences were obtained for both 
selections and replicates* For growth period highly signifi¬ 
cant differences were found for selections but not for repli¬ 
cates* Analysis of variance could not be applied to the protein 
data, since protein determinations were carried out on compo¬ 
site samples from the four replicates. 

















17 



Figure 3 

Partial view of 1944 barley test, showing 
agronomic differences 
between selections 










18 


TABLE I 

Results of the analysis of variance for 
yield and growth period of barley 
selections - 1944 




Mean 

squares 

Variance due to 

D.F. 

Yield 

Growth period 

Selections 

81 

558.3** 

99.7** 

Replicates 

3 

801.6** 

3.7 

Error 

244 

148.0 

6.7 

Total 

328 




** Exceeds the 1 % point. 


Figure 4 shows the relationship between yield in 
bushels per acre and protein content in percent. The slope of 
the regression line is indicated by a bpy value of -.044, a 
highly significant value. The regression equation for protein 
in terms of yield is P » 17.26 - 0.044 y. Thus a gain of 1 % 
in protein Is obtained in this case through a decrease of 22 
bushels per acre in yield. 

Growth period exhibits a definite influence upon the 
relationship between yield and protein content, as Indicated 
by a partial regression coefficient of bpy #m * -*.056. When 
the effect of growth period is eliminated, the regression line 
becomes steeper and a change in protein content of 1% is 
accompanied by a change in yield of only 18 bushels per acre. 









* 

A 

























PROTEIN % 


- 19 - 



YIELD, BU. 


Figure 4 

Relation between yield and protein content 
in the 1944 barley test 






20 - 


The association between yield and protein content 
Is measured by the correlation coefficient r py which has a 
highly significant value of -.533* This compares very closely 
with the results obtained by Waldron (IS)® Protein content 
end growth period were not found to be correlated to a signi¬ 
ficant degree® Yield and growth period, however, gave a highly 
significant correlation coefficient of “*356. This negative 
association bet?/een yield and growth period was hardly 
expected, for this means that the earliest selections were the 
highest in yield® When growth period was held constant, the 
partial correlation coefficient r p ~ m had a value of =>*644® 
Thus, the association between protein content and yield was 
improved by considering it independent of the effect of the 
length of the growth period® 

The multiple correlation coefficient, R p# y- - •764, 
also was highly significant, and proved highly significantly 
greater than r pY ** -.533, when tested by the method illustrated 
by Goulden (5). 

A summary of the regression and correlation coeffi¬ 
cients Is included in Table III® 


1945 

The growing season for 1945 was extremely dry. Only 
4®74 Inches of rain fell during May, June, and July. The 
month of May contributed only 0.26 inches of rainfall to this 































21 


total* All the plots grew well and maintained an erect stand 
which was conducive to both easy and accurate harvesting 
operations* Differences in height* days to heading* and 
growth period cox^ld easily be distinguished* as is shown by 
Figures 5 and 6* 

Results of the analysis of variance for yield* pro¬ 
tein content* and growth period are presented in Table II. 
Highly significant differences were obtained for both selec¬ 
tions and replicates in all three cases. 

TABLE II 

Results of the analysis of variance for 
yield, protein content, and growth 
period of barley selections, 1945 


Variance due to 

D.F. 


Mean squares 

Yield 

Protein 

Growth period 

Selections 

99 

514.06** 

3.80** 

146.5** 

Replicates 

5 

4851.70** 

4.83** 

188.0** 

Error 

297 

57.73 

0.33 

2.7 

Total 

399 





** Exceeds the 1% point. 


The relationship between yield and protein content 
for 1945 is shown in Figure 7. Considering the difference in 
precipitation for 1944 and 1945, it Is surprising to note that 
the range of yields for both years Is approximately the same. 







* 


















• ± r.i. 













22 



Figure 5 

Partial view of 1945 barley test, 
showing agronomic differences 
between selections 







































































25 



Figure 6 

Partial view of 1945 barley test, 
showing agronomic differences 
between selections 



















I 




. 








/ 
















- 24 - 



Figure 7 

Relation between yield and protein content 
in the 1945 barley test 



















as would 


The range of protein values is also similar, though, 
be expected, the 1945 range lies slightly higher. l *he slope 
of the regression line is steeper than for 1944, the bpy value 
being -•066 (highly significant), A decrease of 15 bushels 
per acre in yield would in general mean a 1 % increase in pro¬ 
tein content. The regression equation is P = 19.0 - 0.066 y. 
The length of the growth period showed very little effect upon 
the yield-protein relationship. The partial regression co¬ 
efficient, bpy #ffil s= -.064, is not significantly different from 
the simple regression coefficient. The important point is 
that even for years -showing wide climatic differences the 
general relationship of yield and protein content is of 
approximately the same order. 

A closer association between yield and protein con¬ 
tent for 1945 as compared with 1944 is indicated by a higher 
correlation coefficient, rpy =* -.761 (highly significant). 

The results for 1944 are therefore not only substantiated, but 
improved upon, by the results for 1945. This also holds for 
the association between protein content and growth period, 
which in 1944 was negative though not significant, and in 1945 
gave a negative correlation rp^ » -.524, which was highly sig¬ 
nificant. Yield and growth period gave contradictory results 
for the two years, the correlation coefficient for 1945 being 
positive and highly significant, r^ = +.679. This positive 
correlation coefficient was more to be expected than the nega- 



4 


























tive one of the previous year and, considering the almost ideal 
harvesting conditions, is probably the more nearly correct 
figure. The explanation for the 1944 result may lie in the 
fact that the early varieties were harvested while still stand¬ 
ing erect and without undue loss of grain through handling, 
while the late varieties were harvested under adverse conditions 
caused by lodging. In the 1945 cron, on the other hand, all 
the selections were erect and thus were not subject to error 
through mechanical loss of grain at harvest. A further ex¬ 
planation for the negative association between yield and growth 
period in 1944 may be the possibility that the early, erect 
selections had more completely filled, spikes and, consequently, 
gave higher yields. 

When the effect of growth period was eliminated the 
degree to which yield and protein content varied together was not 
improved, but dropped to the value rpy, m = -.649, which was 
still highly significant. This value compares very closely 
to the partial correlation coefficient for 1944, rpy #m = -.644. 

The multiple correlation coefficient also is very 
similar to the 1944 figure with a value of Rp.ym 5=5 *'762 
(highly significant). This value is not significantly higher 
than the simple correlation coefficient, = -.761. 

A summary of the regression and correlation coeffi¬ 
cients is included in Table III. 



















>1 




87 - 


Means of Two Tests 

When the means of yield and protein content for the 
two years were calculated and plotted, the scatter presented 
in Figure 8 was obtained. The regression coefficient, bpy = 
-•074, was highly significant* The correlation coefficient 
was also highly significant, with a value for r D y of “•VS?. 

Interyear 


Figure 9 presents the scatter of 'points which is 
the result of plotting 1945 yield against 1944 protein con¬ 
tent. The slope is far from steep, but the regression coef¬ 
ficient is highly significant, bpy ~ -.OSS. The association 
between yield and protein content is not striking, but the 
general negative trend is apparent. A highly significant cor¬ 
relation coefficient, r-py « -.566, was obtained. Considering 
the wide differences in growing conditions for the two tests 
from which these data were drawn, a correlation coefficient of 
any larger value would not be expected. 

Plotting 1944 yield against 1945 protein content 
gave the scatter of points depicted in Figure 10. Significant 
results were not obtained in this case for either the regression 
or the correlation coefficient. The indication is that ex¬ 
treme differences in precipitation for the two years resulted 
in highly dissimilar responses each year from many of the 














■* 









■ 


PROTEIN 


~ 28 - 



YIELD , BU. 


Figure 8 

Relation between yield and protein content, 
using the means calculated from the 1944 
and 1945 barley tests 









* 





r 






PROTEIN % 


- 29 - 



Figure 9 

Relation between 1945 yield and 
1944 protein content in 
the barley tests 
















I 


- 30 - 



YIELD , BU. 


Figure 10 

Relation between 1944 yield and 
1945 protein content in 
the barley tests- 






























31 - 


selections. The exhibition of this phenomenon does not come 
as a complete surprise, since differential responses to changes 
of climate—particularly available moisture—-have been noticed 
between standard varieties in past years. 

A summary of the regression and correlation coeffi¬ 
cients for all barley data is presented in Table III, 


TABLE III 

Summary of regression and correlation 
coefficients for barley data, 

1944 and 1945 



1944 

1945 

Based on 2-test means 

b py 

D py.m 

r P7 

r 

rP m 

r ym 

K p.ym 

-.044** 

-.066** 

-.074** 

-.056** 

-.533** 

-.064** 

-.761** 

-.737** 

-.170 

-.524** 


-.356** 

+ o679** 


-.644** 

.764** 

-.649*** 

.762** 



Interyear 


1945 yield/1944 protein 1944 yield/1945 protein 

-. 032 ** 

-.366** 


b py 

r py 


-.011 

-.136 










'i:' ■; . ■ 



32 


Wheat 


1944 

Edmonton (Black soil) 

The 1944 wheat yield-protein test shovred very 
heavy growth in all plots, and in general was late in 
maturing. Differences in height, days to heading, and 
growth period could easily be seen. Several heavy rains 
during the growing season resulted in considerable lodg¬ 
ing. Fortunately very little second-growth appeared. 

In the analysis of variance for yield data, 
highly significant differences were shown for selections, 
and significant differences for replicates. Growth 
period data gave highly significant differences for both 
selections and replicates. The results of analysis of 
variance for both these characters are presented in Table 
IV. Composite samples from four replicates were used for 
obtaining the protein results, thus making analysis of 
variance for protein content impossible. 

Figure 11 presents a scatter of 100 points, 
which indicates the relationship between yield in bushels 
per acre and protein content in percent for the wheat 
test at Edmonton in 1944* The regression coefficient Is 
negative and highly significant, bpy = -.053. With a 









33 


change of 19 bushels per acre In yield there is a cor¬ 
responding change in protein content of approximately 1 %, 
The regression equation is P - 15.18 - 0.053 y. 

TABLE IV 

Results of the analysis of variance 
for yield and for growth period 
of wheat selections, 

Edmonton, 1944 


Variance due to 

D.F. 

Mean 

squares 



Yield 

Growth period 

Selections 

99 

754.09** 

6.38** 

Replicates 

3 

447.19* 

74.06** 

Error 

297 

126.49 

1.57 

Total 

5S9 




** Exceeds the point 
* Exceeds the 5$ point. 


The negative and highly significant correlation 
coefficient, Tpy = -.806, is the highest obtained for any 
of the tests carried out. This measure shows a high 
degree of association between the two characters. It 
indicates that in general there is only a very small 
chance of obtaining a high-yield, high-protein selection. 
However, the possibility of obtaining a high-yield, low- 
protein segregate is greatly enhanced. This is a very 
encouraging feature from the point of view of the soft- 













PROTEIN % 


34 



Figure 11 

Relation between yield and protein 
content in the 1944 wheat 
test, Edmonton 










































































wheat breeder, who is naturally interested in high yields, 
and at the same time desires varieties of low protein con¬ 
tent • 

The correlation coefficient for protein content 
and days to maturity is rather surprising in that it is 
the opposite sign to that obtained for the barley tests of 
both years. Its value is +,423 (highly significant). It 
has generally been considered true that early varieties 
show the highest protein content, but this belief is not 
borne out in this case. This result was perhaps inevit¬ 
able with this series since the protein-yield association 
is high and the yield-growth period association is the 
opposite of that usually expected. 

The correlation coefficient for yield and growth 
period is negative and highly significant, - -.480. 

This is in accord with the result obtained in the barley 
test of the same year, but the association of high yield 
with a short growing period is contrary to the general 
expectation. When growth period is held constant the 
partial correlation coefficient is rpy #m = -.758. This 
figure is highly significant, but is slightly lov/er than 
the simple correlation coefficient rp m ~ -.806. 

The multiple correlation coefficient expressing 
the combined association of yield and growth period with 






*r 



36 


protein content i3 Bp^y^ « *806, which is exactly the 
same value as the simple correlation coefficient for yield 
and protein content, No increase in information on the 
association of greatest interest is obtained, therefore, 
by considering growth period data. 

A summary of the regression and correlation 
coefficients for the Edmonton v/heat test in 1944 is in¬ 
cluded in Table VIII. 

Fallis (Gray soil) 

The Fallis test for 1944 was excellent in all 
respects. An ample supply of moisture resulted in a 
heavy stand of grain which showed no tendency to lodge 
throughout the entire season. It was thus possible to 
carry out harvesting operations with the minimum amount 
of mechanical error. 

Since this test was located a considerable dis» 
tance from Edmonton, continuous observation was impossible, 
with the result that notes on growth period were not 
taken. The results of the analysis of variance for yield 
data are presented in Table V. Highly significant dif¬ 
ferences were obtained for both selections and replicates. 

Yield in bushels per acre was plotted against 
protein content in percent, and the resulting scatter is 



♦ 




presented in Figure 12. A steep regression line is indi¬ 
cated by the highly significant regression coefficient, 
hpy ** -.091. In general, an increase in yield of only 
11 bushels per acre results in a drop in protein content 
of It is interesting to compare these figures with 

those obtained for Edmonton where the increase in yield 
must be twice as much to give the same decrease in protein 
content. The explanation Is no doubt found in the rela¬ 
tive scarcity of available nitrogen in the gray wooded 
soils. The regression equation is P = 11.77 » 0.091 y. 


table v 

Results of the analysis of variance 
for yield of wheat selections, 
Fallis, 1944 


Variance due to 

D.F. 

Yield 

mean squares 

Selections 

99 

155.37**' 

Replicates 

3 

1215.32** 

Error 

297 

34.58 

Total 

399 



** Exceeds the \% point. 


The association between yield and protein con¬ 
tent is not as close as for the Edmonton test, but is 
nevertheless highly significant, rpy *".666. A few of 








PROTEIN 


38 - 



Figure 12 

Relation between yield and protein 
content in the 1944 wheat 
test, Fallis 
































39 


the points have been tagged with the numbers of the selec¬ 
tions they represent in order that a comparison may be made 
with the performance of the same varieties at Edmonton* 

All the points for both stations were compared at the 
time these few were fagged and there was found to be a 
striking similarity in the general position of the points* 
Some differences in performance occurred but, in a high 
proportion of the cases, the response of the selection 
was very similar whether grown on black or gray soil. 
Examination of yield and protein data in Table XX will 
substantiate this statement. 


1945 

Edmonton (Black soil) 

The wheat test at Edmonton in 1945 was grown 
under adverse climatic conditions* A very dry spring 
led to uneven germination, the effect of which remained 
evident throughout the entire growing season* No lodg¬ 
ing appeared in any of the plots. 

For each plot data were recorded on yield, pro¬ 
tein content, and growth period. Analysis of variance 
methods were applied to the data in all three cases and 
the results appear in Table VI* Highly significant dif¬ 
ferences were found for selections and replicates in each 


case • 










'T * , 



























40 


TABLE VI 

Results of the analysis of variance for 
yield, protein content, and growth 
period of wheat selections, 
Edmonton, 1945 


Variance due to 

D.F. 


Mean squares 

Yield 

Protein 

Growth period 

Selections 

99 

156.36** 

1 * 07** 

25.2** 

Replicates 

3 

1527.95** 

5,59** 

86.0** 

Error 

297 

61.62 

0.26 

5.5 

Total 

399 





** Exceeds the 1% point 


Yield In bushels per acre is plotted against 
protein content in percent in Figure 13. The bpy value 
is -*.048 (highly significant). This slope compares very 
closely with that obtained in the 1944 test, an increase 
of 21 bushels per acre in yield being accompanied by a 
loss of 1% In protein content. The regression equation 
is P * 16.1 - 0.048 y. The influence of growth period 
upon the yield-protein relationship is indicated by the 
partial regression coefficient, bpy #m = -.058. When 
growth period is held constant the regression line for 
yield and protein content assumes a steeper slope. 

The correlation coefficient for yield and pro¬ 
tein content is negative and highly significant, rpy «= 









PROTEIN % 


*a* 730 



Figure 13 

Relation between yield and protein 
content in the 1945 wheat 
test, Edmonton 










I 


I 






























-.527. The degree of association between the two charac¬ 
ters is not equal to that of either of the tests in 1944, 
but this may be due in some measure to the variation in 
yields caused by adverse seeding and growing conditions. 

In general, the correlation shown by the scatter of points 
is good. 


The positive correlation between protein con¬ 
tent and growth period for 1944 is substantiated by a 
similar figure for 1945, rp m « +.265, This figure is 
highly significant. 

The association between yield and growth period 
in 1945, in line with that obtained for the barley test, 
is contrary in sign to the result for 1944. The correla¬ 
tion coefficient is positive and highly significant, 
rym * +*255. The association of higher yields with the 
later-maturing selections is more to be expected than the 
negative correlation of 1944. 

The association between yield and protein con¬ 
tent is improved when growth period is held constant, as 
evidenced by the value r*py #m * -.658 (highly significant). 

The combined effect of yield and growth period 
on protein content also gives a highly significant figure, 
^p.^rn * *670. This value Is highly significantly greater 
than the simple correlation coefficient, rpy 33 -.527. 

A summary of the regression and correlation co¬ 


efficients is Included in Table VIII 





» 43 


Fallis (Gray soil) 

A comparatively dry growing season resulted in 
a lighter stand of plants than was obtained in 1944, 

The results of analysis of variance for yield 
and for protein content are presented in Table VII. 

Highly significant differences were demonstrated for both 
selections and replicates in each case. No data were 
recorded for growth period, since the location of the 
test made continuous observation impossible. 

TABLE VII 

Results of the analysis of variance for 
yield and for protein content of 
wheat selections, Fallis, 1945 


Variance due to 

D.F. 

Mean 

Yield 

squares 

Protein 

Selections 

99 

91.55** 

1.39** 

Replicates 

3 

225.58** 

39.48** 

Error 

297 

47.19 

0.78 

Total 

399 




** Exceeds the 1% point. 


The scatter of points obtained by plotting 
yield against protein content is presented in Figure 14. 
The regression coefficient, bpy ® -.079, is highly 



















- 44 - 



Figure 14 

Relation between yield and protein 
content in the 1945 wheat 
test, Fallis 

































45 - 


significant. The slope is not as steep as that for the 
1944 test, but is still above any obtained for tests 
grown on Edmonton soil. The belief that a scarcity of 
available nitrogen plays an important part in the yield- 
protein relationship at Fallis is substantiated. The 
regression equation is P = 14,07 - ,079 y. 

The association between yield and protein con¬ 
tent is essentially the same as that obtained at the same 


station in 1944, 
significant, v 


The correlation coefficient is highly 

",641« 


Means of Four Tests 


The mean values for yield and for protein content 
were calculated, using the data from the four wheat tests. 

The relationship bet?/een these mean values is shown in Figure 

15. 

The regression coefficient, bpy » -.072, is highly 
significant. The correlation coefficient is also highly sig¬ 
nificant, rpy * -.750. Environmental effect has been greatly 
reduced by use of this procedure, and the association shown 
by the correlation coefficient is due almost entirely to 
genetical reasons. 

A summary of the regression and correlation coeffi¬ 
cients for the wheat tests at Edmonton and Fallis, in 1944 
and 1945, is presented in Table VIII. 







If 




■ 






- 46 - 



Figure 15 

Relation between yield and protein content, 
using the means calculated from 
the four wheat tests 









47 


TABLE VIII 

A summary of regression and correlation 
coefficients for wheat tests at 
Edmonton and Fallis, in 
1944 and 1945 



1944 

1945 








Based on 


Edmonton 

Fallis 

Edmonton 

Fallis 

4-test means 

kpy 

bpy ,m 

-#053** 

-.091** 

-.048** 

-.079** 

-.072** 

— 

— 

-.050** 

— 

— 

r py 

r pm 

r ym 

pPy-® 

H p.ym 

-•806** 

-.666** 

-.527** 

-.641** 

-.750** 

+.423** 

— 

+.265** 

«■* 


-.480** 

-- 

+.255** 

» es 


-.758** 

— 

-.638** 

-- 


.806** 

« 

.670** 




** Exceeds the 1 % point# 


A summary of yield and protein data for Edmonton and 
Fallis, 1944, together with the mean values for each selection 
calculated from four tests, is presented in Table IX# 









48 


TABLE IX 


Summary showing: jield and protein data 
for Edmonton and Fallis, 1944; 
and mean values calculated 
from four tests 


1944 


Selection 

number 

Edmonton 

Fallis 

Means 

of 4 tests 

Protein Yield 

Yield 
(bu./ac•) 

Protein 

H) . 

Yield 
(bu./ac.) 

Protein 

ill 

07 

84.7 

11.3 

44.0 

8.3 

11.2 

56.2 

21 

88.2 

11.0 

45.0 

8.1 

11.0 

54.8 

68 

70.5 

11.3 

36.7 

8.0 

10.5 

54.8 

46 

66.5 

11.9 

43.4 

8.9 

11.6 

54.7 

51 

74.8 

10.7 

42.9 

7.9 

10.6 

54.6 

58 

81.7 

10.8 

35.7 

7.8 

10.7 

54.6 

87 

92.6 

10.9 

32.9 

8.5 

11.1 

54.2 

95 

75.9 

11.9 

40.7 

8.4 

11.4 

53.4 

32 

74.1 

11.0 

42.7 

8.4 

11.2 

53,3 

27 

78.2 

10.6 

36.3 

8.6 

11.0 

53.2 

08 

70.9 

11.6 

39.8 

3.3 

11.2 

53.1 

39 

75.7 

10.9 

36.8 

7.9 

10.7 

53.1 

37 

92.7 

11.4 

23.7 

3.5 

11.1 

52.9 

81 

64.3 

13.2 

33.6 

7.8 

11.2 

52,8 

12 

69.8 

11.0 

46.2 

8.0 

11.1 

52.7 

11 

70.8 

11.3 

39.5 

7.9 

11.0 

52.6 

18 

63.2 

10.9 

44.5 

7.7 

10.6 

52.6 

67 

73.4 

11.5 

38.4 

9.0 

11.0 

52.4 

09 

58.3 

12.3 

44.7 

7.5 

10.8 

52.3 

45 

70.8 

10.8 

40.2 

8.0 

10.9 

52.2 

92 

73.4 

11.0 

35.4 

7.9 

10.8 

52.1 

41 

82.8 

10.8 

38.6 

8.4 

10.8 

52.0 

48 

70.7 

11.8 

38.5 

8.6 

11.3 

52.0 

71 

72.9 

11.0 

37.8 

8.5 

11.0 

51.8 

93 

77.5 

11.9 

39.3 

8.1 

11.3 

51,5 

53 

68.5 

11.0 

38.3 

8.2 

11.3 

51.4 

76 

56.8 

12.4 

36.1 

8.1 

11.1 

51.4 

42 

63.1 

11.3 

40.2 

8.3 

11.0 

51.2 

30 

51.3 

12.4 

33.0 

7.9 

10.9 

51.1 

83 

77.9 

11.5 

28.7 

7.7 

11.2 

51.0 

64 

75.2 

11.1 

32.6 

8.9 

11.2 

51.0 

44 

73.0 

11.5 

35.4 

8.4 

11.0 

50.9 

84 

68.8 

12.3 

35.0 

9.0 

11.5 

50.5 

00 

68.5 

11.0 

37.0 

8.3 

10.9 

50.5 

56 

77.1 

10.9 

39.5 

8.3 

11.2 

50.5 

61 

56.3 

11.7 

37.0 

7.8 

10.7 

50.4 












Jf- 






T * 


r 





* *r 

* * 




< 


96 

66 

55 

52 

31 

26 

43 

65 

49 

54 

77 

22 

98 

23 

73 

34 

15 

65 

86 

25 

90 

03 

14 

59 

70 

85 

57 

99 

80 

74 

69 

40 

05 

75 

82 

72 

04 


49 


TABLE IX {continued) 


1944 


Edmonton _ Fallis Means 


Yield 
(bu ./sc. ) 

Protein 

a) 

Yield 
(bu./ac *) 

Protein 

of 4 

Protein 

tests 

Yield 

67.3 

11.1 

33.1 

9.3 

11,6 

50,3 

58.4 

12.2 

37.7 

8.1 

10.7 

50.3 

68.1 

11*5 

39.8 

8.3 

11.2 

50*2 

68.0 

12.0 

40.2 

8.1 

11*2 

50.2 

77.2 

10.8 

39.1 

8,0 

1 -L . 0 

50.1 

75.1 

10.6 

38.0 

7.8 

10,9 

50.0 

58.1 

12.2 

36.9 

8.8 

11.0 

49.9 

70.9 

10.6 

35.7 

7.9 

10.5 

49.4 

73.0 

11.7 

53.4 

8,3 

11.3 

49,5 

64.6 

11.8 

32.4 

7.9 

11 • 1 

49,1 

55.2 

11.5 

35.8 

7.6 

10.8 

49.0 

72.8 

10.8 

35.1 

8.3 

11.0 

48,9 

66.5 

11.4 

39,6 

8,3 

11.2 

48.8 

61.7 

12.2 

36.7 

8.6 

11.6 

48,8 

62.5 

11.5 

34.1 

8,4 

10.8 

48,7 

67.4 

11.1 

36.8 

8.0 

10.8 

48.7 

71.0 

12,2 

37.4 

8.5 

11,7 

48.6 

71.1 

11.4 

62 • * 4 : 

8.5 

10.9 

48.6 

50.8 

12.4 

47.3 

8.4 

11,6 

48.5 

57.3 

12.4 

35.5 

8.1 

11.4 

48.2 

63.7 

11.9 

35,5 

8.1 

11.3 

47,8 

62.5 

11.2 

36.6 

8.1 

10,7 

47.8 

67.1 

11.4 

32.9 

8.6 

11.4 

47.7 

55.9 

12.0 

42.6 

8.1 

11.2 

47.6 

62«4 

11.3 

39.9 

8.7 

11.1 

47.4 

63.1 

11.2 

34.1 

7.8 

11.0 

47,2 

57.2 

12.2 

41 oO 

8.0 

11.3 

47.2 

64.4 

12.4 

32.5 

8.0 

11.3 

47.0 

67.3 

11.4 

31.9 

9.0 

11.1 

46.7 

63.8 

11.6 

35.1 

8.6 

11.4 

46.6 

51.3 

12.9 

31.1 

9.5 

11.6 

46.6 

45.0 

13.2 

36.8 

8.2 

11.1 

46.2 

50.6 

11.6 

31.0 

8.8 

10.9 

46.2 

54.6 

12.8 

30.5 

9.5 

12.0 

45*6 

64.6 

11.8 

50.9 

8.7 

11.8 

45.3 

64.0 

11.2 

30.9 ‘ 

8.5 

11.2 

45.1 

54.8 

12.8 

29.0 

8.5 

11.3 

45.0 

44.6 

13.0 

38.3 

8.4 

11.6 

45.0 

67.0 

11.8 

30.2 

8.9 

11.6 

45.0 

48.2 

13.2 

32.4 

9.3 

11. s 

44.8 










TABLE 


IX (continued} 


Selection 

number 


1944 


Means 

of 4 tests 

Protein Yield 

Edmonton 

Fa 

Ills 

Yield 
(bu./ac . ) 

Protein 

ill 

Yield 

(bu./ac. 

Protein 

) a) 

16 

55.9 

12.2 

31.4 

8.5 

11.8 

44.0 

29 

57.6 

12.2 

29.4 

8.3 

X J_ e t; 

45,8 

79 

41.7 

13.8 

31.7 

9.7 

12*1 

42.9 

IS 

46.5 

11.3 

32.7 

8.6 

11 * 4 

42.8 

36 

59.0 

12.9 

34.2 

9.5 

11.5 

42.6 

20 

58.2 

11.6 

23.7 

8.5 

11.3 

41.8 

33 

50.1 

11.4 

26.1 

8.8 

11.4 

41,4 

97 

53.2 

12.2 

26.8 

9.1 

11.8 

41.2 

17 

39.6 

13.5 

25.7 

10.9 

12.6 

40.8 

10 

45.4 

12.1 

32.0 

8.8 

11.8 

40.6 

24 

38.9 

13.5 

28.5 

10.5 

12.1 

39.6 

13 

36.3 

12.4 

25.6 

9.7 

11.2 

39.5 

02 

46*2 

13.7 

27.2 

10.1 

12.5 

39.2 

50 

44.3 

12.5 

32.8 

9.0 

11.9 

38.8 

88 

49.6 

12.9 

24.0 

10.2 

12.1 

38.1 

62 

36.8 

12.2 

22.8 

9.0 

11.4 

37.8 

94 

49.8 

12.4 

35.6 

9.0 

11.7 

37.7 

06 

33.4 

13.6 

28.3 

10.4 

12.4 

37.5 

91 

36.8 

13.9 

26.5 

9,5 

12,4 

37.3 

01 

38.0 

14.0 

24.8 

11.1 

12.8 

36.6 

89 

39.6 

14.4 

24.7 

9.8 

12.5 

35.5 

38 

45 • 0 

13.1 

25 .3 

10.1 

12.3 

55.4 

23 

42.8 

12.6 

24.2 

S.7 

12.0 

34.9 

35 

27.0 

14.4 

20.0 

11.5 

13.2 

30.3 


Interstation 

Examination of the tagged selections in Figures 11 
and 12 indicates the tendency for t.a© same selections to per¬ 
form in the 3 a mo manner at different stations* <won though 
the Edmonton plot 3 were placed on black soil and the Fallis 













« 



*• 





l 


51 


plots on gray soil, the points representing the various selec¬ 
tions remained in the same general position. 

The effect of soil environment was still further 
minimized by plotting the Edmonton yield against the Fallis 
protein for 1944, The regression coefficient in this case was 
rather low, but still highly significant, b py = -,036, The 
more important point is the fact that this combination of data 
from two stations still gave a highly significant negative 
correlation coefficient, r py * ->*627, The correlation surface 
is presented in Figure 16, A test for non-linearity of the 
regression line was applied to the data, using the method out¬ 
lined by Goulden (5), The regression line proved to be non¬ 
linear. The linear regression depicted in Figure 16, therefore, 
is not entirely correct. 

The corresponding but reverse combination of Fallis 
yield and Edmonton protein for 1944 is plotted in Figure 17. 

As would be expected, these results also were highly signifi¬ 
cant. The regression coefficient is high, owing to a wide 
range of protein values and the narrow range of yields, bpy = 
-.086. The association between yield and protein content is 
good, being negative and highly significant, r py = -.539. 

The same procedure was carried out using the 1945 
data, and gave similar results. When Edmonton yields were 
plotted against Fallis proteins, the scatter presented in 
Figure 18 was obtained. The regression coefficient is bp y » 
-.049 (highly significant). The correlation coefficient is 


*r 































* 




. 







PROTEIN % 


- 52 - 



Figure 16 

Relation between Edmonton yield and 
Fallis protein content in the 
1944 wheat tests 




















PROTEIN % 


- 53 - 



Figure 17 

Relation between Fallis yield and 
Edmonton protein content in 
the 1944 wheat tests 























PROTEIN % 


- 54 » 



Figure 18 

Relation between Edmonton yield and 
Fallis protein content in 
the 1945 wheat tests 


















55 - 


also highly significant, r py « -,468. The combination of 
Pallis yields with Edmonton proteins also gave highly signifi- 
cant results. The scatter of points is shown in Figure 19, 

The regression coefficient is b py ~ -=*,054, and the correlation 


coefficient is r p y - -,494, 


Interyear 


For the Edmonton wheat test, 1944 yields were plotted 
against protein values for 1945. The resulting scatter 
(Figure 20) shows no significant trend. As would be expected, 
the combination of 1945 yield and 1944 protein content gives a 
similar result. The scatter is shown in Figure 21. The cor¬ 
relation coefficients, respectively, are r p y » -.097 and rpy ** 
-.052, neither of which is significant. The inference is that 
the various selections exhibited different responses to the 
highly dissimilar moisture conditions of the two years. This 
lends support to the results obtained with interyear combina¬ 
tions of barley data. 

Interyear combinations of Fallis data, on the other 
hand, show a definite trend. When 1944 yields are plotted 
against 1945 proteins, the scatter presented in Figure 22 is 
obtained. In this scatter the trend is not obvious, but a 
regression coefficient of -.022 and a correlation coefficient 
of -.218 were calculated. These figures are significant only 
to the b% point. 



H 










* 














* 







- 56 - 



Figure 19 

Relation between Fallis yield and Edmonton 
protein content in the 1945 wheat tests 























- 57 




I 

O 

Q: 

Q- 



YIELD, BU 


Figure 20 

Relation between 1944 yield and 
1945 protein content in the 
Edmonton wheat tests 













PROTEIN % 


- 53 





YIELD , BU. 


Figure 21 

Relation between 1945 yield and 
1944 protein content in the 
Edmonton wheat tests 









- 59 - 



Figure 22 

Relation between 1944 yield and 
1945 protein content in the 
Fallis wheat tests 























•> 












- 60 


The corresponding combination of 1945 yields with 
1944 proteins gives highly significant results. The scatter 
ol points is shown in Figure 23• The regression coefficient 
here is very high, bpy ~ =*,103, a change in yield of only 10 
bushels being accompanied by a change in protein content of 
1/a. The association between yield and protein content is also 
negative and highly significant, * -.616. 

A summary of interstation and interyear results is 
presented in Table X. 


TABLE X 

A summary of regression and correlation 
coefficients for interstation and 
interyear combinations of 
yield and protein data 



b py 

r 

1 py 

Interstation 



1944: Edmonton yield/Pallis protein 
Fallis yield/Edmonton protein 

-.036** 

-•086** 

-.627** 

-.539** 

1945: Edmonton yield/Pallis protein 
Pallis yield/Edmonton protein 

-.049** 

-.054** 

-.468** 
-.494** 

Interyear 



Edmonton: 1944 yield/1945 protein 

1945 yield/1944 protein 

-.004 
+ . 008 

-.097 
+ .052 

Pallis: 1944 yield/1945 protein 

1945 yield/1944 protein 

-.022* 

-.103** 

-.218# 

-.616** 


* Exceeds the 5 % point 

** Exceeds the 1 % point. 











61 



Figure 23 

Relation between 1945 yield and 
1944 protein content in the 
Fallis wheat tests 





















62 


DISCUSSION 


It must be concluded from the foregoing results that 
a definite negative relationship exists between yield and pro¬ 
tein content in wheat and barley* In each test, with no 
exception, the association was negative and highly significant. 
The fact that the simple correlation coefficients were highly 
significant in all six tests, two of which were carried out 
using barley selections, and four using soft wheat selections 
grown for two years on both black and gray soil, indicates that 
the association is, to a large extent, genetically controlled. 
Environment undoubtedly plays a part in the yield-protein 
relationship, but when fche effect of environment is reduced by 
the use of interstation results, or by calculations based on 
mean values, the high negative correlation remains. Examina¬ 
tion of the correlation coefficients based on the general means 
indicates that for both wheat and barley approximately 35 percent 
of the genetical potentiality for protein content is determined 
by yielding ability. Some type of genetical linkage is Indi¬ 
cated for yield end protein content. 

Interstation results show that the various selections 
give similar responses on the two soil types. This emphasizes 
the belief that the relationship between the two characters is 
largely genetically controlled, and that the characters are in 
s one manne r linked. 



. 

. 










■ • • 




63 


Interyear results are not as conclusive as those for 
interstation data* This leads to the belief that the ability 
to produce protein may be physiologically tied up with other 
factors* Thus, in two such widely different growing seasons 
as those in 1944 and 1945, the physiological balance may have 
been, changed, leading to dissimilar responses from many of the 
selections during the two years. It is interesting to note 
that for interyear data for Fallis, where the relative scarcity 
of available nitrogen is more a limiting factor than the 
physiological ability of the plant to make use of it, the 
response of each selection over the two years is more nearly 
alike then at Edmonton, where nitrogen is not a limiting fac¬ 
tor of stich importance* 

The slope of the regression line is sufficiently 
steep in each case to render the negative relationship 
economically important* In breeding for wheat of high pro¬ 
tein content, if a regression coefficient of -.05 can be 
generally expected, an increase in protein content of 1 % means 
a loss in yield of 20 bushels per acre. Regression coeffi¬ 
cients based on Pallia data are particularly high. Introduc¬ 
tion of higher-yielding varieties of wheat to the gray soil 
zone may therefore result in even lower protein values than 
those now obtained* 

The length of the growing period apparently exerts 
some influence on yielding ability* Conflicting results were 







■(. 






























64 


obtained for 1944 and 1945, The reason for these differences 
may be found in the precipitation records for the two years. 

In May, June, and July of 1944 over 11 inches of rain fell. 
During the corresponding period in 1945 the precipitation 
amounted to only 4.75 inches, with only 0.25 inches in the 
crucial month of May. The negative relationship for 1944 may 
be partially explained by the fact that under the influence of 
heavy rains the later varieties tended to lodge. The lodged 
spikes failed to fill as completely as those standing erect, 
thus leading to a decreased yield for these selections. In 
1945, the selections which filled the best were those which 
were late in maturing, and which were therefore able to take 
some advantage of the late rains• 

The length of the growth period also exerts some 
influence on the relationship between yield and protein con** 
tent. This is Indicated by the fact that in two tests the 
slope of the regression line was made steeper by holding the 
effect of the growth period constant. The partial correlation 
coefficient was also found to be greater in these t?/o cases 
than the simple correlation coefficient. The multiple cor¬ 
relation coefficient, which expresses the combined effect of 
yield and growth period on protein content, was found to be 
highly significantly greater than the simple coefficient. 

It would appear, from the foregoing statements, that 
the possibility of obtaining varieties which are high in both 


• ' 

■ . 

- 

♦ 

■' ■’ ' •' 

. r ' J 

. 



65 


yield and protein content is fairly remote. Such selections 
have occasionally been made, but the chance of obtaining them 
through selection work based mainly on agronomic characters 
is made much more difficult by the negative relationship which 
has now been demonstrated. This is of particular interest to 
the breeder of hard spring wheats and feed barleys. On the 
other hand, the breeder of soft wheats and malting barleys 
will experience little difficulty in obtaining the combina¬ 
tion of high yield and low protein content which he desires. 

It may be concluded that protein content can serve 
as a valuable criterion in the selection or■discarding of 
certain strains from early-generation hybrids of wheat and 
barley. Protein determinations could be made on all strains 
which had been selected on the basis of their agronomic charac¬ 
teristics from head-row material. In the case of hard spring 
wheats, keeping the negative yield-protein relationship in 
mind, approximately 15$ of the selections showing the highest 
protein content could be discarded, since their yielding 
ability would in all probability be very low. Approximately 
15$ of the selections having the lowest protein content could 
be discarded for this characteristic alone, ^or malting 
barley or soft wheat, approximately 35$ of the selections 
characterized by high protein content could be discarded, 
since both high protein content and low yielding ability are 
undesirable features. A considerable saving would thus be 


‘ 


. ' 


- 66 


made In time, labor, and land, which would otherwise be used 
in bringing these strains up to the yield-test stage. 

SUMMARY 


In order to determine the relationship between 
yield and protein content of random selections from, single 
crosses in wheat and barley an experiment was devised which 
reduced to a minimum the error due to different varieties and 
different environments. For each of wheat and barley a cross 
was made between a high-yield-low-protein parent and a low- 
yield-high-protein parent. After the progeny of these crosses 
had segregated for several generations, 100 plants were selected 
at random from each crop. Each lot of seed was numbered and 
increased during the following year. In 1944 and in 1945 
these selections were grown at Edmonton In replicated yield 
tests. The wheat selections were also grown on the gray soil 
at Fallis. Records were kept of yield, protein content, and 
growth period for each selection. 

The results of analysis of variance showed highly 
significant differences between selections for yield, protein 
content, and growth period* A highly significant negative 
relationship was demonstrated between yield and protein con¬ 
tent in all six tests. The regression of yield on protein 
content was of such magnitude as to make this association 









— 

* 






















67 


economically important. The effect of growth period on the 
yield-protein relationship was not constant enough to justify 
definite conclusions* 

Correlation coefficients \?ere calculated for both 
barley and wheat, using the mean values for protein content 
and yield* A highly significant negative-relationship was 
demonstrated. Combinations of data from different stations and 
from different years were also used. Interstation results 
gave highly significant negative correlation coefficients* 
Interyear results showed the same general trend, though the 
results were not as conclusive* Methods such as these prac¬ 
tically eliminate the effect of environment, and indicate that 
the yield-protein relationship Is largely genetically con¬ 
trolled* 

This negative relationship greatly decreases the 
chance of obtaining a high-yield-high-protein variety of 
either wheat or barley* The problem of selecting strains of 
soft wheat or malting barley is greatly simplified. 

With a negative yield-protein relationship estab¬ 
lished, protein determinations carried out on head-row samples 
could serve as a valuable criterion for selection or discard¬ 
ing of many strains. 





3 

























68 


ACKNOWLEDGEMENTS 


The writer wishes to express his sincere gratitude 
to Dr, A, G. McGalla, head of the Department of Plant Science, 
for his suggestion of the problem and for friendly counsel 
during the progress , of the ?/ork. The assistance given by 
student members of the department staff is deeply appreciated. 
During the latter half of the research period the 
writer was the holder of a University of Alberta Research 
Scholarship, 


REFERENCES 


1. ANDERSON, J.A., MEREDITH, W.Q.S., and SALLANS, H.R. 

Malting quality of Canadian barleys, IV, A summary 
of information of special interest to plant breeders. 
Sci. Agr. 23:297-314. 1943. 

2. CLARK, J.A. Breeding wheat for high protein content. Jour. 

Amer. Soc. Agron. 18:648-661. 1926. 

3* ____ and QUISENBERRY, K.S. Inheritance of yield and 

protein content in crosses of Marquis and Kota spring 
wheats grown in Montana. Jour, Agr. Res. 38:205-217. 
1929. 

4. COX, G.M. and ECKHARDT, R.C. The analysis of lattice and 

triple lattice experiments in corn varietal tests. 

Iowa State Coll, of Agr. Res. Bull. 281:1-65. 1940. 

5. GOULDEN, C. H. Methods of Statistical Analysis. John 

Wiley and Sons, N. Y. 1939. 













I 


6* GOtiLDEN, G • H. and ELDERS, A •!» A statistical study of* 

the characters of wheat varieties influencing yield. 
Sci. Agr. 6:357-345. 1926, 

7* HARRINGTON* J.B, Predicting the value of a cross from an 
Pg analysis. Gan. Jour. Res. 6:21-37, 1932. 

8* .. Yielding capacity of wheat crosses as 

indicated by bulk hybrid tests. Can. Jour, Res, C.18: 
578-584. 1940. 

9. HAYES, H.K. and IMMER, P.R. Methods of Plant Breeding. 

McGraw-Hill Book Co., Inc. Hew York and London* 1942. 

10. IMMER, F.R. Relation between yielding ability and hoznozy- 
gosis in barley crosses. Jour. Amer. Soc. Agron. 
33:200-206. 1941. 

11* JENKINS, B.C. The relations of stoiaatal size and generation 
to the yielding ability of bulked wheat hybrids. M.Sc. 
thesis, Univ. of Alta, 1944. 

12. MALLOCH, J.G. and NEWTON, R. The relation between yield and 

protein content of wheat. Can. Jour. Res. 10:774-779. 
1934. 

13. McCALLA, A.G. and ROSE, D. Quality of Alberta-grown wheat. 

Alberta Agr. Ext. Bull. 37:1-36. 1941. 

14. MEREDITH, W.O.S. and OLSON, P.J. Cultural studies with 

barley. IV. Summary of results for yield and malting 
quality. Sci. Agr. 23:237-246, 1942. 

15. NEATBY, K.W. and McCALLA, A.G. Correlation between yield 

and protein content of wheat and barley in relation to 
breeding. Can. Jour. Res. 0,16:1-15® 1938. 

16. OLSON, P.J., MEREDITH, W.O.S., LAIDLAW, H.C., and LEJEUNE, 

A.J. Cultural studies with barley. I. Differential 
responses of varieties to date of seeding with 
respect to yield. Sci. Agr. 22:225-241. 1941. 

17. TAYLOR, D.K. A study of parental lines and bulk progenies 

from single and compound crosses in wheat and barley. 
M.Sc. Thesis. tJniv. of Alta. 1943. 

18. WALDRON, L.R. Yield and protein content of hard red spring 

wheat under conditions of high temperature and low 
moisture. Jour. Agr. Res. 47:129-147. 1933. 



+ 





+• 



19. WHITESIDE, A.G-.O. Statistical significance of wheat protein 

percentage differences in varietal trials. Can. Jour. 
Res. 0,14:387-393. 1936. 

20. WORZELLA, W.W. Some objectives in breeding for yield and 

other agronomic characters in wheat. Airier. Soc. 

Agron. 33:174-180. 1941. 




































& ■ 


sS 



















B29752